Automated Hyper-Parameter Optimization for Eye Movement Artifact Removal

Brain-computer interface (BCI) systems allow users to control external devices with their brain waves. However, electroencephalography (EEG) signals used by most BCI systems are prone to artifacts from various sources (e.g., muscle activity, eye movements, and electrical interference). These artifacts can degrade the performance and usability of BCI systems. Many tools exist to eliminate these artifacts. However, not all methods are automated, and some might require tuning certain hyper-parameters for optimal performance. We propose a method to automatically optimize the hyper-parameters of an eye blink artifact removal tool to improve the removal of artifacts in resting state EEG. We use a subset of eye movement artifacts to optimize the hyper-parameters using the EEG Quality Index (EQI) as the objective function. The optimized hyper-parameters are then used in a test artifact to quantify the improvement of the EQI. Results show improvement in the EQI when compared to the default artifact removal hyper-parameters, and raw EEG traces. We conclude that our method can provide a personalized and robust artifact removal solution for BCI users with complex needs.